Strengthening analyses of line-up procedures: a log-linear model framework

作     者：Luby, Amanda S. 

作者机构：Carnegie Mellon Univ Dept Stat Pittsburgh PA 15213 USA 

出 版 物：《LAW PROBABILITY & RISK》 (Law Probab. Risk)

年 卷 期：2017年第16卷第4期

页      面：241-257页

核心收录：

学科分类：0301[法学-法学] 03[法学] 0714[理学-统计学（可授理学、经济学学位）] 0701[理学-数学] 

主　　题：eyewitness identification line-ups ROC log-linear models 

摘      要：Over the past three decades, there has been considerable interest among both researchers and the criminal justice system to reform line-up procedures to ensure eyewitness identifications are as accurate as possible. Recently, the Receiver Operating Characteristic (ROC) methodology has been adopted to analyse line-up procedures, but it has not been universally accepted in the field. This article examines the application of ROC methodology to line-up data and proposes an approach based on log-linear models as an alternative method to analyse line-up procedures. We find that log-linear models allow for non-binary classification schemes that are necessary for analysing line-up outcomes. Conditional independence relationships between variables can also be identified through a log-linear model. Log-linear models also provide a natural framework to account for multiple sources of uncertainty present in eyewitness identification data. While the ROC analysis provides valuable insight into the changes in outcomes across different decision-making thresholds, incorporating a log-linear analysis allows us to examine these outcomes in finer detail.